Time-variant spectral analysis of LDF signals on the basis of multivariate autoregressive modelling

Time-variant AR-modelling was used to analyse the non-stationary properties of LDF signals during provocation tests. The procedure for the estimation of time-varying AR parameters based on Kalman filtering is presented. The estimates can be used to calculate instantaneous measures, such as peak frequency, spectral band power and coherence. The resulting course of instantaneous peak frequency of a sinusoidal signal with frequency jump was compared to similar parameters derived from short-time FFT and Hilbert transformation. Univariate time-variant spectral analysis was used to investigate LDF measurements in patients with Raynaud's phenomena. The experimental protocol was splitted into phases of different room temperature. The results demonstrate time-dependent variations of spectral components (amplitude and frequency). By means of time-variant coherence analysis of LDF and diameter measurements of vessels in a hamster skin fold the existence of a main rhythm around 0.1 Hz in the LDF signals which is related to vasomotion is shown.